跑通Multi-Agent-Constrained-Policy-Optimisation

✅ 1. 克隆项目并创建 Conda 环境

git clone https://github.com/chauncygu/Multi-Agent-Constrained-Policy-Optimisation.git
cd Multi-Agent-Constrained-Policy-Optimisation
conda create --prefix ./envs python=3.7 -y
conda activate ./envs

---

✅ 2. 安装 PyTorch 和依赖

pip install torch==1.10.1+cu111 torchvision==0.11.2+cu111 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
pip install -r requirements.txt

---

✅ 3. 安装系统依赖项（OpenMPI、GLEW、OSMesa、Patchelf）

sudo apt update
sudo apt install -y libopenmpi-dev openmpi-bin libglew-dev libosmesa6-dev patchelfpip install -r requirements.txt

---

✅ 4. 安装 MuJoCo 2.1

sudo apt install ntp -y
sudo ntpd -gqsudo apt update
sudo apt install --reinstall ca-certificateswget https://mujoco.org/download/mujoco210-linux-x86_64.tar.gztar -zxvf mujoco210-linux-x86_64.tar.gzexport MUJOCO_PY_MUJOCO_PATH=$(pwd)/mujoco210
export LD_LIBRARY_PATH=$MUJOCO_PY_MUJOCO_PATH/bin:/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATHpip install -U 'mujoco-py<2.2,>=2.1'

---

✅ 5. 解决 libstdc++.so.6 兼容问题（如需）

ENV_PATH=$(python -c "import sys; print(sys.prefix)")
ln -sf /usr/lib/x86_64-linux-gnu/libstdc++.so.6 $ENV_PATH/lib/libstdc++.so.6

---

✅ 6. 安装 MACPO 和 MAPPO-Lagrangian 模块

cd MACPO
pip install -e .
cd ../MAPPO-Lagrangian
pip install -e .

---

✅ 7. 验证 MuJoCo 安装是否成功

cd ..
python -c "import mujoco_py"

---

✅ 8. 运行训练脚本

cd MACPO/macpo/scripts
chmod +x ./train_mujoco.sh
./train_mujoco.sh

---

✅ 9. 额外依赖项（如提示缺失）

pip install jinja2